GWAS come under a lot of criticism. Some of the common ones that I hear are:

  • Most of the effects are tiny, what do we do with this information
  • Most of the hits are in non-coding regions of the genome, how do these lead to a phenotype
  • GWAS significant hits only explain a small portion of the heritability of the trait

I wondered if there’s a GWAS that might satisfy some of these critics. I think I found one. Here’s this paper by Eimear Kenny et al. in Science 2012. The phenotype is striking and super interesting, blond hair in Solomon Islanders. The sample size of this study is 85 individuals. This population has the largest proportion of blonde hair outside Europe(5-10%). The top hit is a coding mutation in a gene called TYRP1, a melanosmal enzyme. All of the signal at this loci can be attributed to this coding polymorphism, when the effect of this polymorphism is accounted for the other hits at this locus are no-longer significant. This loci along with some simple covariates such as sex and age explains ~ 46% of the phenotypic variance.

This study is one of the earlier studies which shows how GWAS in diverse populations can uncover hits that are impossible to detect in Europeans(the predominant study population used in GWAS to date.) I find this study fascinating. The reason this GWAS is so close to perfect is due to the study design. The authors identified an interesting phenotype in an isolated population and then searched for the genetics of this trait. The trait, blond hair, makes sense for a genetic study and is unlikely to be hugely influenced by the environment, unlike a trait like say diabetes which could have a huge amount of non-genetic variance. In this case the trait had a smoking gun, a coding variant in a gene that makes sense with respect to the biology.

A lot of the genetic studies today advocate for 100s of thousands and even millions of participants. This particular study is so clever that it was able to find a hit using just 85 individuals that would not be found using millions of participants. We can argue that this phenotype is an outlier but I think there are lessons in study design here that we need to carefully think about. Sequencing everything that we can get our hands on and then figuring out why is certainly wasteful.

I have no doubt that there’s much untapped signal like this in the rest of the world. Sometimes I want to catch a flight to India and start recruiting study participants to uncover some fascinating genetics.